During a conversation with an interactive system, an avatar is typically used to provide responses to a user. The avatar is a computer-generated image, which represents a person in any suitable format in a virtual environment. However, most avatars used during such conversations are same for most users and provide same responses without personalization or taking into account interests of the user. Such a situation is usually not suitable as the user may need to talk to different persons at different points of time in different scenarios during the conversation. In an existing implementation, the avatar used is a fixed avatar and is always independent of a conversation. In another existing implementation, the avatar works on pre-stored emotions. Such implementations, however, are unable to provide a rich and fulfilling user experience during the conversation.